A Multi-Modal Framework Combining Proteomics, Electrophysiology, and Clinical Phenotypes to Characterize Neuroplasticity in Parkinson’s Disease Rehabilitation

Background: Rehabilitation for Parkinson’s disease (PD) is limited by the absence of reliable biomarkers to monitor neuroplasticity and predict treatment response. Objective: To investigate how proteomic and electrophysiological signatures relate to clinical recovery and to evaluate whether multimodal integration enhances prediction of rehabilitation outcomes. Methods: In this retrospective cohort of 165 PD patients, participants received either intensive multimodal rehabilitation (IMR, n=110) or standard rehabilitation (SR, n=55). Circulating neurotrophic and inflammatory proteins (BDNF, GDNF, VEGF, IL-6, TNF-α) and electrophysiological measures (EEG spectral power, TMS-evoked MEP) were assessed at baseline and post-intervention, alongside clinical scales (UPDRS-III, gait). Multivariate regression and predictive modeling were applied. Results: IMR was associated with significant increases in neurotrophic factors and reductions in inflammatory markers, paralleled by EEG β desynchronization, γ enhancement, and MEP facilitation. In multivariate analyses, ΔBDNF and Δβ power independently predicted motor improvements, while ΔMEP amplitude was the strongest predictor of gait recovery. A multimodal model integrating proteomic and electrophysiological features achieved superior discrimination for achieving the minimal clinically important difference (AUC=0.74) compared with single-modality models. Conclusion: Multimodal integration of proteomic and electrophysiological markers provides complementary insight into neuroplastic mechanisms and improves prediction of rehabilitation response in PD. These findings support the development of biomarker-driven strategies to personalize neurorehabilitation.